Indexes, just like relational database!

Week 4 on M101J - Indexes

• Indexes take addtional space, but provide much faster data retrieval.
• Creating indexes
• .ensureIndex( )
• e.g. "db.students.ensureIndex( { student_id : 1 } );
• e.g. "db.students.ensureIndex( { student_id : 1, "class" : -1 } );
• Multi-key index
• index key can be array
• e.g. db.bbb.insert( { a: [1, 2, 3], b: 1 } );
• keys cannot both be arrays
• e.g. db.bbb.insert( { a: [1, 2, 3], b: [4, 5, 6] } );
• Unique index
• index key has to be unique, duplicate not allowed
• e.g. db.students.ensureIndex( { student_id: 1, name: 1 }, {unique: true} );
• Foreground vs background indexing
• Foreground: fast, but blocks writes. Suitable for DBA. Can lock a replica while it's being indexed.
• Background: slow (2~4x slower), concurrent with writes. Suitable for developers in a production setting.
• .explained( )
• Useful to examine a query to see if indexing is utlilized
• Important keys: "cursor" (did it use BtreeCursor?), "nscannedObjects" (how many objects actually queried?)
• .hint( )
• $natural: returns result in its natural order
• .hint( { $natural: 1 } ) will use BasicCursor instead of BTreeCursor

Nice course from 10gen

10gen (maker and distributor of MongoDB) offers some very nice courses on MongoDB:
https://education.10gen.com/‎

Notes from M101J - week3:

Cool stuff about MongoDB schema

• Rich documents
• Store array of data
• Pre-join data (embed data)
• Fast access
• No "Mongo Joins"
• No constraints
• No primary key / foreign key
• Atomic transaction operation
• Within one document
• No declared schema
• Similar structure in documents

Living without transactions

• Atomic operation
• In order to accomplish it:
• restructure code to work within same document.
• Implement locking mechanism / semaphore
• Tolerate inconsistency
• One to one relationship
• Embed or not to embed depends on:
• Freq of access
• sSize of items ( > 16MB? )
• Atomicity of data
• Benefits of embedding
• Improved read performance
• One round trip to DB
• High latency: 1ms
• High bandwidth
• "Write" latency can be sig. improved by embedding data
• Decision to denormalize
• 1:1 - Embed
• 1: many - Embed (from many to 1)
• many : many - Link (using array of _id)

Multi-dimentional skills needed

After two weeks into all things MongoDB, this is what I think needed to be an expert:

• Setup, monitor, and administer MongoDB on servers.
• Understand / use MongoDB in application development.
• Troubleshoot issues as they arise.
• Database migration, from other dbs to MongoDB
• "Sharding" - understand deeper scope, when does it occur and how?
• Distinguish differences between MongoDB and other dbs, pros / cons.
• Understand how MongoDB performs / reacts on different storage technologies (SAS vs SSD vs PCI).
• Understand advanced inner-working of MongoDB.